Naive Bayes
Naive Bayes assumes conditional independence of features and classifies via Bayes' theorem, P(y|x)∝P(y)∏P(x_i|y); it is simple and efficient.
Purpose
To provide a fast baseline classifier for text classification and high-dimensional data.
Mechanism
Bayes' theorem plus the conditional independence assumption: P(y|x)∝P(y)∏P(x_i|y). Training estimates the prior P(y) and the conditional probabilities P(x_i|y). Gaussian/multinomial/Bernoulli NB suit different data types.
Use cases
- Text classification
- Spam filtering
- Sentiment analysis